Multispectral Fine-Grained Classification of Blackgrass in Wheat and Barley Crops

• DOI: 10.48550/arxiv.2405.02218
• Publication Date: 2024-05-03
• Authors: Madeleine Darbyshire; Shaun R. Coutts; Eleanor Hammond; Fazilet Gokbudak; Cengiz Öztireli; Petra Bosilj; Junfeng Gao; Elizabeth Sklar; Simon Parsons
• Keywords: FOS: Computer and information sciences; Computer Vision and Pattern Recognition (cs.CV); Computer Science - Computer Vision and Pattern Recognition

As the burden of herbicide resistance grows and environmental repercussions excessive use become clear, new ways managing weed populations are needed. This is particularly true for cereal crops, like wheat barley, that staple food crops occupy a globally significant portion agricultural land. Even small improvements in management practices across these major worldwide would yield considerable benefits both environment global security. Blackgrass grass which causes particular problems north-west Europe, production area, because it has high levels well adapted to agronomic practice this region. With machine vision multispectral imaging, we investigate effectiveness state-of-the-art methods identify blackgrass barley crops. part work, provide large dataset with evaluate several key aspects recognition. Firstly, determine performance different CNN transformer-based architectures on images from unseen fields. Secondly, demonstrate role spectral bands have classification. Lastly, size classification each models trialled. We find even fairly modest quantity training data an accuracy almost 90% can be achieved